Researcher, interested in autonomous machine learning, reinforcement learning, robotics, 3d printing.... Professor at Uni Tübingen and Co-founder of Ontic Labs
Joined July 2014
- Tweets1.2K
- Following209
- Followers2.2K
- Likes2.9K
Pinned Tweet
I will present our paper on diffentiable simulation at #ICLR in the first poster session: P3-#1304
How can we make gradient-based optimization work for contact rich and stiff dynamics?
Our answer: adaptive stepsizes and contact form distances plus some contact routine smoothing
Georg Martius retweeted
It’s great to see growing interest in MBRL. Last year in RWM, we demonstrated that policies trained entirely in world-model imagination transfer zero-shot to the real world.
Sometimes, ideas are ahead of their time. Glad to see this direction getting the recognition it deserves.
🧠Model-Based RL shows promises but has seen limited success in real-world robotics.
🌎Introducing Robotic World Model, a black-box end-to-end neural dynamics model that bridges this gap, where policies are trained purely in imagination.
@NeurIPSConf
🎯sites.google.com/view/roboti…
Georg Martius retweeted
Excited to share these news! Was awesome to work with the folks @AIatMeta, the speed at which they are building is crazy. Looking forward to what is next!
Georg Martius retweeted
🇺🇦/🇷🇺/🇺🇸 Je résume le génie du cretin de la Maison Blanche :
- Il retire les sanctions contre le pétrole russe pour toucher l’équivalent de 3 jours de consommation aux USA.
- Ça envoie un signal de normalisation à la Russie, et lui permet de toucher de l’argent pour financer sa guerre.
- Les européens et la délégation ukrainienne aux USA au même moment passent pour des idiots en découvrant l’annonce.
- Trump appelle Zelensky à quitter son poste de Président, l’insulte, demande à l’Ukraine d’arrêter d’attaquer le nerf de la guerre russe.
- Zelensky accepte, à condition que la Russie arrête elle aussi les attaques contre l’infrastructure énergétique.
- On apprend dans la presse que Trump menace de couper les renseignements à l’Ukraine, mais aussi de mettre un terme au programme PURL qui permet aux europeens d’acheter des armes américaines pour l’Ukraine.
- Sans consulter ni Moscou ni Kyiv, Trump annonce une pseudo trêve énergétique entre les deux belligérants.
- Moscou découvre l’annonce.
- Kyiv découvre l’annonce.
- Zelensky s’engage à accepter une trêve si la Russie elle aussi s’y engage.
- Moins de 2h plus tard, la Russie attaque une centrale énergétique à Kharkiv.
Trump est vraiment le plus grand guignol et lâche de la politique internationale.
Une honte pour les États-Unis.
Un incapable qui maquille ses échecs aux dépens de ceux qui résistent.
Une ordure qui se couche devant les pires régimes.
Une humiliation quotidienne pour les pays occidentaux.
J’espère qu’il prendra la plus grosse raclée possible aux élections de mi-mandat.
Georg Martius retweeted
🤖 How much can a humanoid learn from human demonstrations?
Check out our new paper: VioLA, Learning Generalist Humanoid Control Policies from Human Data.
Georg Martius retweeted
Our paper got accepted to #NeurIPS2026 🎉🥳.
If you are intersted in why the brain or AI systems might benefit from complex neurons, check it out arxiv.org/abs/2605.12049, it's really cool :)
Huge thanks to @GMartius and @SelfOrgAnna and all the support from the lab 🙏
1) I'm excited to present "Scaling Laws and Tradeoffs in Recurrent Network of Expressive Neurons" 🎉, where we question the optimality of simple neurons for network performance, and capture scaling behaviours using a minimal information theoretic model: arXiv.org/abs/2605.12049
Georg Martius retweeted
I mean, possibly yes, but what is the importance of having a lot of knowledge that nobody understands?
I am not against AI, but maybe we should reconsider what is important in academia. Publishing papers is not important, understanding new knowledge is.
Georg Martius retweeted
Evals first progress is the way!! Glad this is public now!
Georg Martius retweeted
Walking up stairs vs sprinting on slopes, which to learn first? When task difficulty isn’t obvious, handcrafting curriculum becomes challenging.
We tackle this with automatic curriculum and achieve the highest rough-terrain speed on ANYmal.
🔗sites.google.com/view/lp-acr…
🧵RAL 2026
Georg Martius retweeted
I wonder given all those particles fits, can we distill/itentify the params in the sim like Newton so that we don't have to stick to learned module
Excited to share our #CoRL2026 work PhysCoRe: a physics-corrected world model for deformable objects!
We couple a physics-based simulator with two feed-forward modules: one infers material from vision, the other corrects its internal dynamics.
Check: lunarlab-gatech.github.io/Ph…
Georg Martius retweeted
📢 The CORE COGNITION lab (core-cognition.github.io) is looking to hire another PhD student, working on automated scientific discovery and knowledge representation in cognitive science.
Starting date: March to September 2027.
🙏 Please share widely!
Georg Martius retweeted
The DINO team at FAIR Paris is hiring a PhD student to push the frontier of visual representation learning. We offer hard research topics, a research and engineering team supporting you, and a stimulating PhD community. Help us shape the next generation of DINO models! Details👇
Georg Martius retweeted
✈️ Announcing: The CoRL 2026 Travel Support Program for students & early-career researchers, with:
🔹 Free registration
🔹 Hotel (shared, paid by organizers)
🔹 $500 travel stipend
🗓️ Apply by 5 Oct (AoE)
corl.org/attending/inclusion…
Please share! 🙏
#CoRL2026 #RobotLearning
Georg Martius retweeted
A nice statement from the mathematicians: mathandai.org
Georg Martius retweeted
Build a humanoid robot yourself! 🪚
OpenArm is an open-source humanoid robot.
It comes with full CAD, control code, firmware, and simulation tools, everything needed to build, modify, and operate it.
The arms are designed to be compliant and backdrivable. Teleoperation is supported, with force feedback and real-time gravity compensation so operators can guide the arm naturally.
Super important, on the simulation side, OpenArm works with platforms like MuJoCo and Isaac Sim, letting developers test policies in virtual environments before running on hardware.
Assemble it yourself from a kit or get it prebuilt, the goal is accessibility for research labs, small teams, and enthusiasts.
The project is run by @enactic_ai in Tokyo, Japan, and aims to lower the barrier for experimenting with dexterous manipulation.
Let's put robotics in mainstream! 🔥🔥
🔗 Find the link here: github.com/enactic/OpenArm
~~
♻️ Join the weekly robotics newsletter, and never miss any news → ziegler.substack.com
Georg Martius retweeted
Más del 60% del hábitat del orangután de Borneo (Indonesia) fue destruido solo en los últimos 3 MESES en incendios provocados para que los capitalistas sigan instalando más plantaciones de aceite de palma.
Nadie habla de este ecocidio provocado por intereses del capitalismo, los orangutanes están siendo captados vagando hambrientos buscando algo de comida fuera de su hábitat, han destruido todo su entorno solo para hacer negocio.
Es una situación crítica la que se vive en Borneo y es el perfecto reflejo de este sistema depredador que arrasa con todo solo por aumentar los beneficios económicos de las multinacionales, que meten el aceote de palma en el 50% de los productos del supermercado.
Georg Martius retweeted
Excited to introduce FB-MEBE, led by @jhu486 ! A Behavior Foundation Model (BFM)🤖 for quadrupedal control trained without rewards AND without pre-collected data! Given a reward at test-time, it recovers an optimal policy zero-shot and transfers seamlessly to real hardware.🧵
Georg Martius retweeted
The Rhine is disappearing before our eyes. Record low levels in parts of Germany & the Netherlands as the intense heatwave continues. This isn't just a river story. It's a climate story. As the planet warms, heatwaves are pushing agriculture and water supplies to breaking point.
Georg Martius retweeted
Bis zu 12 Teams können an dieser SPRIND Challenge teilnehmen - und erhalten in den nächsten 24 Monaten bis zu 1,5 Mio. Euro. Die Bewerbungsfrist läuft bis zum 16.10. Um diese SPRIND Challenge vorzustellen, bieten wir am 20.08. ein Webinar an buff.ly/0QisO2b
Georg Martius retweeted
🚀 We are thrilled to welcome Dr. Kashyap Chitta to the ELLIS Institute Tübingen as our newest Principal Investigator and Hector-Endowed Fellow!
Kashyap is a leading expert in autonomous systems and physical AI. At the Institute, he will launch a new research group, the Physical AI Lab, focused on robot learning in both simulation and the real world.
This approach will reduce the cost of physical interaction, enabling deployment of physical intelligence to a wide range of robotic systems and environments.
Alongside this academic role, he will continue his work with @kesai_labs, a non-profit open-science AI lab he co-founded in May 2026.
Please join us in giving him a warm welcome! 👏
#ArtificialIntelligence #AutonomousDriving #ComputerVision #EmbodiedAI #MachineLearning #ELLISTübingen @Cyber_Valley
Georg Martius retweeted
Ontic Labs is Live on X 🚀 and we are hiring!
We are one of 10 teams selected for SPRIND’s €125M Next Frontier AI Challenge to build Europe’s next frontier AI labs.
Our mission is to build foundation models for physical intelligence that learn from human video: both how the physical world behaves and how tasks are carried out inside it.
Our goal is to enable robots to draw on the breadth of human experience and act dexterously, safely, and efficiently across tasks, environments, and embodiments.
Our team:
@omidtaherii, @mikel_zhobro, @wielandbr, @GMartius, @GerardPonsMoll1
Want to build with us? We are hiring exceptional talent in world modeling and foundation models.
More details and opportunities:
onticlabs.io
Supported by @SPRIND
@MPI_IS @uni_tue @ELLISInst_Tue @tubingen @Cyber_Valley
#NextFrontierAI #PhysicalAI #WorldModels #Robotics #Hiring
Photo: Felix Adler für SPRIND
